Krovak EPSG translation wrong?

[warmerdam]

Frank,

not sure where this belongs to (mailing list):

epsg_tr.py -proj4 2065
# S-JTSK (Ferro) / Krovak
<2065> +proj=krovak +lat_0=49.5 +lon_0=0 +alpha=30.28813972222222
+k=0.9999 +x_0=0 +y_0=0 +ellps=bessel +pm=ferro +units=m +no_defs  <>

-> lon_0=0 looks wrong to me

grep '<2065' /usr/local/share/proj/epsg
<2065> +proj=krovak +lat_0=49.5 +lon_0=42.5 +alpha=30.28813972222222
+k=0.9999 +x_0=0 +y_0=0 +ellps=bessel +pm=ferro +units=m +no_defs  <>

Is this a bug somewhere?

thanks Markus

[warmerdam]

I have confirmed that EPSG:2065 is being expanded by OGR with a lon_0=0 value (including in the WKT) currently. I'm getting the same results as far back as GDAL 1.4 though I'm not positive I was really using different csv files. I'm not absolutely sure whether the lon_0=42.5 Markus found in an old epsg init file is correct or not, but the matter certainly bears additional investigation.

[martinl]

Replying to warmerdam:

I have confirmed that EPSG:2065 is being expanded by OGR with a lon_0=0 value (including in the WKT) currently. I'm getting the same results as far back as GDAL 1.4 though I'm not positive I was really using different csv files. I'm not absolutely sure whether the lon_0=42.5 Markus found in an old epsg init file is correct or not, but the matter certainly bears additional investigation.

The reference meridian used in Krovak projections is 42d30m (prime meridian is Ferro), so

lon_0=42.5 +pm=ferro

is right.

[Even Rouault]

I can reproduce the problem when running GDAL svn trunk against $(GDAL_ROOT)/data directory from svn trunk. But when running against /data directory from GDAL 1.5 or 1.4, I get the expected value of lon_0=42.5.

Extract of pcs.csv from gdal svn trunk:

PARAMETER_CODE_7	PARAMETER_VALUE_7	PARAMETER_UOM_7
8833			42.3			9110

Extract of pcs.csv from gdal 1.5 branch:

PARAMETER_CODE_2	PARAMETER_VALUE_2	PARAMETER_UOM_2
8812			42.3			9110

8812 == ProjCenterLong 8833 == PolarLongOrigin --> that looks wrong.

Last upgrade to EPSG 6.17 in r14987 seems to be the root of the problem

[Even Rouault]

This affects GDAL trunk (1.6.0dev)

[warmerdam]

I'm guessing that EPSG has changed how the central meridian is communicated for the krovak projection, and ogr_fromepsg.cpp will need to be updated to handle both approaches. I'll dig into this.

[warmerdam]

I have corrected ogr_fromepsg.cpp (ImportFromEPSG()) logic in trunk (r15479) to accept either the ProjCenterLong or PolarLongOrigin parameters. This appears to have corrected the problem.

Thanks for the digging Even.